基于关键帧标识的视频人脸识别技术研究  

Face Recognition Technology Research on Video Keyframe Identification

作　　者：徐辉[1] 凌捷[1] 

机构地区：[1]广东工业大学计算机学院,广东广州510006

出　　处：《计算机技术与发展》2012年第9期80-82,86,共4页Computer Technology and Development

基　　金：广东省自然科学基金项目(9151009001000043);广东省教育部产学研合作项目(2011A090200068)

摘　　要：针对人脸检测与跟踪问题提出在监控视频中面向复杂背景、多姿态的快速人脸检测。使用了基于Haar特征,并结合有效的预处理手段跟搜索策略,另外在检测算法的基础上,考虑到监控视频的序列的特点,结合背景差、肤色、前帧参考等因素,提出一套基于帧间差分法来识别视频中人脸内容变动并标识此关键帧的方法,并以视频日志的形式记录下来。实验证明,此方法可以在针对检测特定目标提高监控视频人脸识别的正确率和检测速度。Aiming at face detection and tracking issues, the multi-profile fast face detection is proposed in the surveillance video in the complex background. Use the Haar-based features,combined with an effective means of pre-search strategy,in addition to the detection algorithm based, taking into account the characteristics of surveillance video sequences, combined with background subtraction, skin col- or,the fast reference frame and other factors,a set of method is presented based on inter-frame difference method to identify the changes of human face in the video and identify the keyframe, and record in the form of video log. Experiments show that this method can im- prove the accuracy and detection speed of face recognition of video surveillance for detecting specific target.

关 键 词：关键帧标识 视频日志 人脸识别 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]